Detecting counterfeit 4110 mosfets

[methods]

Just as an aside -

I agree with all that was said about mosfets (even though I know jack about them ). I would like to say that IR may be an exception to the rule when it comes to stated specifications.

At my work we do extremely high consequence builds where every single component is qualified to a 20 year life - It can be nearly impossible to qualify most components.
International Rectifier (IR, as in IRFB4110) has such a good reputation that they are almost rubber stamped.
If IR states that their mosfet is rated for 100V then you can take that to the bank if you keep all other parameters within specification.
My elder co-workers tell me that IR has underrated their parts for years in order to be the turn-key source for government contracts.

I have no fear each time I turn on my controller at 100.8V and run it hard to 200A peaks.
I dont think that anyone needs to tip-toe around their 4110 modified controllers or worry about fatigue... I can grantee that any one of the 50 other Chinese sourced parts will fail long before these mosfets will...

I cant say that I would feel as bold with the knock-off 4110's though.

-methods

 

[liveforphysics]

I want to see the 4110 get the voltage turned up across it until it has a failure. I bet it takes 130v.

 

[methods]

I can supply more fets for this test.

-methods

 

[Doctorbass]

Here is some interesting "real world" data:

old analog crystalyte 12 fets with the true IRFB4110:

used between 100.8V and 84Vdc at up to 100A ( 30A avg) and often 70A burst)

I also towed a schoolbus with the X5 and this controller ( the motor phase was nor changing very fast cause the wheel was like bloked when towing the 19200lbs schoolbus.. so the fets was stressed alot for high current!!

still work perfect for one years and half!!!...

Doc

 

[John in CR]

My pack of 80 real 4110's arrive today, yay! Thanks Methods. My current 15 fet infineon controllers have unknown fets, since the #'s are sanded off, but the controllers have always run hotter than the motor even with the controllers very well ventilated, and that's running at the stock 60v30a.

Can I expect my controllers to run much cooler with the high quality 4110's? Does most of the waste heat of a controller come from the FETs?

John

 

[Doctorbass]

My pack of 80 real 4110's arrive today, yay! Thanks Methods. My current 15 fet infineon controllers have unknown fets, since the #'s are sanded off, but the controllers have always run hotter than the motor even with the controllers very well ventilated, and that's running at the stock 60v30a.

Can I expect my controllers to run much cooler with the high quality 4110's? Does most of the waste heat of a controller come from the FETs?

John

i'm pretty sire your controlelr will run moch cooler since the 4110 have very low Rds On resistance.. = less heat dissipation.

Also.. I would say that at high power( kW +) the major part of the heat will come from the mosfet.. but at low power.. the heat may come more from the regulator big resistance.

What do you think Camllight?

 

[BatteryMooch]

Ooh, I get to use my favorite line..."It depends".

A MOSFET might have a lower on-state resistance (Rds-on), which can allow it to run cooler at the same current level, but its gate capacitance might be higher. This would cause the MOSFET to spend more time in its linear region as it was turned on and off and that would cause more heating.

Also, different MOSFETs can have different junction-to-case and case-to-sink thermal resistances. This can have a HUGE effect on their temperatures.

So, the answer is...it depends.

 

[John in CR]

Ooh, I get to use my favorite line..."It depends".

A MOSFET might have a lower on-state resistance (Rds-on), which can allow it to run cooler at the same current level, but its gate capacitance might be higher. This would cause the MOSFET to spend more time in its linear region as it was turned on and off and that would cause more heating.

Also, different MOSFETs can have different junction-to-case and case-to-sink thermal resistances. This can have a HUGE effect on their temperatures.

So, the answer is...it depends.

BUT, I'm going from unknown fets that result in what seems to me to be unusually hot controllers to proven high quality fets that people report nice cool controllers at higher power than I currently run, so wouldn't there have to be some kind of controller design problem causing all this heat for it not to run cooler with the 4110's?

John

 

[liveforphysics]

John-

If I had to take a wild-ass-guess, I bet you can expect 1/2 to 1/3 of the heat on your controller. It will also mean less voltage drop, which means more of the energy from your pack will get turned into useful mechanical energy.

For a TO220 format FET, the 4110 has a very significant advantage over generic FETs. It's as good as it gets for a FET in that package size. Pretty badass little chip.

 

[BatteryMooch]

BUT, I'm going from unknown fets that result in what seems to me to be unusually hot controllers to proven high quality fets that people report nice cool controllers at higher power than I currently run, so wouldn't there have to be some kind of controller design problem causing all this heat for it not to run cooler with the 4110's?

John

In your case, I think you're probably right. The MOSFET drivers would have to be pretty darn wimpy to handle what I think will be the higher gate capacitance of the 4110's compared to the ones you have now.

My answer is still "it depends", and it still does. But I think you're pretty safe. I'd probably make the mods if I was in the same situation. I just wanted to point out that even a great MOSFET like the 4110 can be problematic in some situations and isn't the perfect choice to blindly select for all situations.

 

[bhavinfirst]

I am delighted to find this forum.
From the discussion above i get the impression that some of you are using fets at max volt rating. I always kept atleast 30-50% margin for the spikes when switching inductive loads. Am i over estimating the spikes
Also, are you'll just replacing the fets on ready controller or making one? Any1 who has made 100A controller, i bow down 2 him. stray inductance and not so good layout is already causing me troubles @ 40A
During testing @ 100V, is switching being done with inductive load??bcoz thats what makes all the difference

 

[BatteryMooch]

Welcome to the ES forums!

Personally, I keep about a 30% voltage margin as I just can't be sure that I'm seeing all the spikes on my scope and I don't mind spending a little more money for the higher-rated FET. But, I'm designing stuff from scratch and am not limited by what a controller might be able to deal with (when replacing FETs).

Unfortunately, no one can predict if you're overestimating the voltage rating you need because every controller is different. If every aspect of your controller's operation was understood (max spike levels, etc.), I'd still have at least a 20% margin as the farther you operate from the max voltage rating, the longer your FETs will operate without problems. Statistically, that is. There's always a small chance that anyone's FETs can blow at any time, no matter how timidly they're run.

 

[amberwolf]

Based on my experiences, I agree with the margin of Vds for safety. I had a crappy ScootNGo brushed motor controller that could barely qualify *as* an electronic speed controller that was already blown up when I got it. Once I had poked thru the charred mess around the MOSFETs and driver transistors and gate resistors to figure out what actually originally connected to where, wired across the evaporated traces and gaps in the burned-away PCB, I tried a lot of random MOSFETs (in identical pairs) out of various junked electronics I had laying around (since my projects' goals include using all-recycled parts wherever possible).

First I looked up the specs on them, to see what I could even use on the 24V controller, and then used anything that was rated for that voltage or higher. In the end, even stuff up to 40V and even 50V blew up, not from overheating or overcurrent either.

Some of it could just be that they had been micro-damaged in their previous lives, but even given that I toasted a LOT of MOSFETs. Even a couple of freewheeling diode modules were destroyed in the process.

Not having a scope at the time I had no way to prove if the inductive spikes or RF from the brushes might be what killed them, but that's what I suspect.

To make sure I had enough current to charge the gate I'd used some 4A 40V Motorola JFETs I had a bag of PNP and another of NPN, and those never died.

I got my hands on some NTY100N10's with 100V Vds and 123A Ids max, and used a pair of those on there, and I never had a MOSFET failure after that, even after upping the voltage to 36V (3x 12V 12A SLA), no matter what motor I put on there, from the little scooter motor (well, what was left of it) to some radiator fan motors, to some much better radiator fan motors in pairs, to a treadmill motor, to a wheelchair motor driving a 120 pound bike!

They even stayed very cool, despite having only about 7 or 8 square inches of total heatsink area, if that, with ALL of the heat-producing parts bolted on the same heatsink (gate drivers, fw diodes, and MOSFETs).

But I think that voltage headroom is what really let me keep working them, rather than any other characteristic.

Now I'm using a totally different controller (the 2QD by http://4QD.co.uk) running some Fairchild 60V 80A 3.8mOhm, FDP038AN06A0, in pairs per leg of half-bridge, and probably due to the lower RDSon, they're cool with a much smaller heatsink, only about 4 or 5 square inches total including the little fins (out of an old PC power supply). I used at least 8 or 9 square inches per MOSFET with the NTY100N10s for the same result (but could probably have used half that), with only one NTY100N10 per leg.

So another caveat seems to be that sometimes using the biggest MOSFET you can find doesn't give you the best results!

My controller is now considerably smaller with the FDP038AN06A0s than the NTY100N10s, even though I am using twice as many of them. I have not quantified this, but I think I get better motor performance from it, too, because there is less controller resistance wasting voltage that could otherwise drive the motor. It "feels" better when riding, if that makes any sense.

 

[gip_mad]

Hey Cam, any updates? Did you get my mosfets? I really need to order a bunch of 4110 from methods as I have burned the hell through my stash of crappy mosfets developing my controller!

 

[ejonesss]

is there any performance differences other than heat.

meaning will a controller with the fake fets produce less power and even act like a car taking off in high gear?

 

[BatteryMooch]

RdsOn! I want to see the difference in heating between the real and fake 4110s

Ask and ye shall receive. Well, eventually.

I finally finished up a couple of projects and had a chance to run some tests. This was a Rds-on (drain-to-source resistance, when on) comparison between two batches/sources of fake IRFB4110 MOSFETs and two batches of genuine IRFB4110 MOSFETs. This test compares the amount of heat a MOSFET will generate while it's fully turned on (for a specific amount of current). It does not address heating due to switching losses, that is, heating that occurs while the MOSFET turns on and off. There are some very interesting results!

Fake IRFB4110
Body markings:
FB4110
1F 7P
P645J
tiny center recessed area had "MALAY" text in it

10A current, starting Rds-on values (in milliohms):
19.9
18.0
18.0
17.5
19.0
16.5
15.5
17.3
17.2

10A current, Rds-on values after temperature settled several minutes later to around 126C on rear face of MOSFET, approx, 176C junction temperature. In same order as above.
35.4
34.4
34.2
34.0
35.5
32.8
28.0
36.6
34.3


Fake IRFB4110
Body markings:
FB4110
1F 7P
P645J
tiny center recessed area had no text in it, just a rough-textured surface

10A current, starting Rds-on values (in milliohms):
8.1
9.0
7.5
7.7
11.7
7.9
11.7
9.2
7.7
7.3

10A current, Rds-on values after temperature settled several minutes later, temperature not recorded. In same order as above.
9.6
10.7
8.7
9.1
15.6
9.2
15.6
11.3
8.8
8.4


Genuine IRFB4110
Body markings:
IRFB4110
920P
S6 UR
tiny center recessed area had 2 characters of text in it, not recorded.

10A current, starting Rds-on values (in milliohms):
4.5
4.5
4.6
4.6

20A current, starting Rds-on values (in milliohms):
9.5
9.0
9.2
9.2

10A current, Rds-on values after temperature settled several minutes later, MOSFET was barely above room temp. In same order as above.
4.8
4.7
4.8
4.9

20A current, Rds-on values after temperature settled several minutes later, approx 86C case temp. In same order as above.
14.7
14.3
14.1
14.7


Genuine IRFB4110
Body markings:
IRFB4110
928P
5V KY
tiny center recessed area had 2 characters of text in it, not recorded.

10A current, starting Rds-on values (in milliohms):
4.6
4.7
4.7
4.7
4.9
4.8
4.8
4.8
4.9

20A current, starting Rds-on values (in milliohms):
9.2
9.3
9.4
9.3
9.8
9.8
9.6
9.5
9.8

10A current, Rds-on values after temperature settled several minutes later, MOSFET was barely above room temp. In same order as above.
4.9
5.0
5.0
5.2
5.1
5.1
5.0
5.2

20A current, Rds-on values after temperature settled several minutes later, approx 86C case temp. In same order as above.
15.1
15.2
15.3
15.4
15.4
16.2
15.6
15.0
15.2


Conclusions and comments:
The genuine MOSFETs had significantly lower, and much more consistent, Rds-on values compared to the fakes. For the genuine MOSFETs, the results were a good match to the MOSFET's datasheet specs. Always a good thing. In fact, the genuine MOSFETs ran so much cooler than the fakes I was able to double the amount of current flowing through them and still have them run 40C cooler than one batch of the fakes! These results cannot be directly translated to the thermal environment the MOSFETs will see in a controller but they give you a great idea of how much better the genuine MOSFETs are.

IMHO, the little bit of money you save by buying counterfeit MOSFETs is a total waste. The genuine MOSFETs are worth the cost.

[Edit] I forgot to mention that the genuine 4110's had significantly thicker (heavier gauge) legs than the fakes.


Test conditions:
- 300A, 5V regulated power supply as source of test current, set to 5.00V, supply was given a 2 hour warm-up before testing started.
- Fluke 8846A benchtop DMM reading Rds-on, digital filtering on to average any ambient electrical noise, all readings rounded to nearest 0.1mV, meter was given a 2 hour warm-up before testing started.
- CamLight CC-400 constant-current electronic load set to 10A (actually 10.05A) and 20A (actually 20.03A). Worst case drift was less than 0.15% from these starting values.
- All MOSFET gates driven directly from a battery pack. Voltage ranged from 12.56V-12.54V during the tests.
- All MOSFETs mounted by the tips of their legs, vertically, in free air at 22C-25C ambient. No heat sinks were used. No fan or obviously moving air was present during the tests.
- The drain and source connections were made by 12AWG stranded cable, mechanically clamped to the legs 1/8" from the body of the MOSFET. As a check of the sensitivity of the cable clamping position on the leg, I move the cable out to the end of the leg, over 1/2" from the body. The Rds-on value only increased by an average of 2%. Experimenting with clamp pressure only change the Rds-on values less than 1%. I'm confident that any variations in my cable-to-leg clamping from MOSFET to MOSFET had a negligible effect on the readings.
- MOSFET case temperatures measured on center of the "rear" face (side that presses against a heat sink). It's the only place to get an accurate enough reading to use the thermal resistance specs to derive the junction temperature.

 

[liveforphysics]

Thank you so much Camlight!

It would seem a 6-fet with genuine IRF4110's is a more capable controller than a 12-fet with fake IRF4110's.

 

[methods]

Excellent.
Now I have something to point to when people ask me "But are they really better.....?"


Now lets talk about switching losses.
Since this testing started some time back I have had the chance to do a lot more testing.
I am going to go out on a limb and say that most overheating issues with quality controllers (say genuine 4110) is due to switching losses and not due to raw rdson.
(yea - I am sure you guys knew that but I had to prove it to myself )

I first started noticing this in my early testing with the 100V 100A controllers. While testing them at 100V 30A the controller would get quite warm. Similar tests at 100V 100A would result in almost no heating. It seems that if we want to rate a specific controller to a specific power level we must take into consideration how well the KV (and ultimately the current limit) of the motor is matched to the system. Since most riders want to go out and buy the "fastest" (translated highest KV) motor they can find and then pair it up with the cheapest controller (translated low current limit) they can find systems are often times very poorly optimized.

So.. When I make recommendations to people about 6 fet controllers and I say that we effectively want to run them with no DC current limit (though - the phase limit is set to control peak stall currents) it is very hard for people to understand what I am driving at -

If a system is balanced so as that you can run a controller with no DC current limit it will run *cooler* than another (lower power) system that is bouncing off the current limit because we are running the mosfets in a more efficient way - less time in PWM.

More torque, less heat.
Win Win!

But.... Then as a few of us 5305 owners have learned (and Luke and Fechter have pointed out to me) if you go too far in the low KV direction the internal resistance of the motor becomes the true limiting factor.....

As Luke suggested via IM the other day - I am starting to think that I might be able to make more power by running a slightly higher KV (lower I^2*R loss) motor paired with a higher current limit controller. I started doing some simulations to prove / disprove this and ended up getting busy with another project.

Anyhow - Thanks for the tests!

Just yesterday I received 3,300 Genuine IRFB4110 fets from IRTronix. Check my sig if you need any.

-methods

 

[BatteryMooch]

Thanks for the MOSFET donation for the tests! And a big Thank You to everyone else who donated MOSFETs too. Being able to test two different batches/manufacturers of the fake ones really helped show how variable the resistance is for these.

Methods, would it help you if I did a turn-on/turn-off timing test of a couple from each batch, genuine and fake (to help with calculating switching losses)?

 

[methods]

Nah- you have done enough

-methods

 

[whatever]

Just to continue a bit on this thread
One thing that I dont think is pointed out clearly ( unless I missed it)
is that the P after the date stamp is identifying as from the mexico plant.
Thats a very good indicator of quality if that 'P' is there ( assuming it does actually mean mexico site)
It would be good to know more about the other codes, such as lot code, and the mystery XX idents ( in red below)



Methods mentions two things:
that

Purchased direct from the factory

If methods can confirm he went to tijuana factory that would pretty much seal the "P" as indicating mexico site ( or was it via irtronix which isn't factory direct)
You can purchase the genuine fets in china also at good prices through irf china ( they get them supplied from mexico)
Another methods quote:

Turns out the TO-220 is ONLY built in Mexico - nowhere else

If that is true, then the 'P' is a very good indicator of quality.
Interestingly I"ve been testing some batches of fets which are both from the "p' factory, both irf4310 type,
with identical markings except for the "xx" mystery ident, and they show different values from a simple resistance test across the drain and source.
I suspect that even with the same factory, that depending on which line the fets are made on there are differences in quality. Purely based on these 4310 fets.
More info on the assembly lot code and xx code might be useful

 

[BatteryMooch]

Just to continue a bit on this thread
One thing that I dont think is pointed out clearly ( unless I missed it)
is that the P after the date stamp is identifying as from the mexico plant.
Thats a very good indicator of quality if that 'P' is there ( assuming it does actually mean mexico site)

Isn't that only true until the counterfeits start adding a "P"?
That is, if they haven't already and a lot of FETs we think are genuine actually aren't.

IIRC, the IR web site has reliability data on a lot of their FETS and I think the lot numbers are included.

 

[liveforphysics]

It always confuses me with counterfeiter's of mosfets. If they have the technology to make semiconductors, why on earth can't they load the correct graphic to etch the chip with? Seems like if I was going to copy fets, making it be worded and appear physically identical would be the most simple part of trying to copy a semiconductor. lol

 

[whatever]

as far as i'm aware, there aren't any fets cloning the irf etchings, as lforphysices says. The clones are easy to identify at this stage, if you have the correct etchings and a 'P' its mexico source ( at this stage), it is truly bizarre indeed.
The mystery 'xx' numbers, I checked a number of fets from same batch, they have all different numbers in that circle, even in same batch, I think that number may not be much use ( who knows!), it will say something particular about the fet, if anyone else can check their 'p' mexico fets, I think will find the 'xx' is not the same even within one batch.

 

[BatteryMooch]

I checked the IR web site and it appears that, unless Mexico is the only lead-free line for IR, that a "P" only means lead free.
http://www.irf.com/package/marking/pm_to220ab.pdf